Harvestが少ないサンプル数で動作しない件

Harvest_Sample.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2018-01-28T23:34:07.697989Z",
     "start_time": "2018-01-28T23:34:07.688821Z"
    }
   },
   "outputs": [],
   "source": [
    "import world4py\n",
    "\n",
    "world4py._WORLD_LIBRARY_PATH = '/Users/yamachu/.anyenv/envs/pyenv/versions/anaconda3-4.3.1/lib/python3.6/site-packages/world4py/libworld.dylib'"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2018-01-28T23:34:08.323488Z",
     "start_time": "2018-01-28T23:34:08.108810Z"
    }
   },
   "outputs": [],
   "source": [
    "from world4py.np import apis, tools, structures"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2018-01-28T23:34:08.332717Z",
     "start_time": "2018-01-28T23:34:08.326787Z"
    }
   },
   "outputs": [],
   "source": [
    "x, fs, nbit = tools.get_wave_parameters('./demo/sample.wav')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2018-01-28T23:34:08.370992Z",
     "start_time": "2018-01-28T23:34:08.336913Z"
    }
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(array([  0.00000000e+00,   0.00000000e+00,   0.00000000e+00, ...,\n",
       "          0.00000000e+00,   0.00000000e+00,   3.05175781e-05]), 16000, 16)"
      ]
     },
     "execution_count": 5,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "x, fs, nbit"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2018-01-28T23:34:08.395165Z",
     "start_time": "2018-01-28T23:34:08.386219Z"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Help on function harvest in module world4py.np.apis:\n",
      "\n",
      "harvest(x, fs, f0_floor=71.0, f0_ceil=800.0, frame_period=5.0, **dummy)\n",
      "    F0 extract by Harvest\n",
      "    \n",
      "    Args:\n",
      "        x (ndarray(dtype=double, ndim=1)): Input waveform\n",
      "        fs (int): Sampling frequency [Hz]\n",
      "        f0_floor (double, optional): Floor of F0 estimation\n",
      "        f0_ceil (double, optional): Ceil of F0 estimation\n",
      "        frame_period (double, optional): Frame shift [ms]\n",
      "    \n",
      "    Returns:\n",
      "        ndarray(dtype=double, ndim=1): extracted F0\n",
      "        ndarray(dtype=double, ndim=1): Temporal positions\n",
      "    \n",
      "    Notice:\n",
      "        Harvest tend to consider F0 as voiced compared to Dio\n",
      "\n"
     ]
    }
   ],
   "source": [
    "help(apis.harvest)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2018-01-28T23:34:08.497691Z",
     "start_time": "2018-01-28T23:34:08.454609Z"
    }
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(array([   0.        ,  483.68521641,  477.32521021,  462.67253647,\n",
       "         431.64963385,  386.43432658,  344.71813637,  315.4458332 ,\n",
       "         294.01678507,  279.44260727,  276.99089551,  288.12248287,\n",
       "         308.25729845,  327.89784622,  341.95233607,  350.21905512,\n",
       "         353.63351189,  353.41241359,  349.85006721,  344.08769285,    0.        ]),\n",
       " array([ 0.   ,  0.005,  0.01 ,  0.015,  0.02 ,  0.025,  0.03 ,  0.035,\n",
       "         0.04 ,  0.045,  0.05 ,  0.055,  0.06 ,  0.065,  0.07 ,  0.075,\n",
       "         0.08 ,  0.085,  0.09 ,  0.095,  0.1  ]))"
      ]
     },
     "execution_count": 7,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "apis.harvest(x[0:1600], fs, frame_period=5.0)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "サンプル数が1410ぐらいになるとセグフォで落ちる  \n",
    "=> 大体100msecぐらいは最低でもHarvestには必要そう"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "ExecuteTime": {
     "start_time": "2018-01-28T23:34:23.652Z"
    }
   },
   "outputs": [],
   "source": [
    "apis.harvest(x[0:1410], fs, frame_period=5.0)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.6.3"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}

正直100msecあってもF0の抽出誤りが起きまくるからなんともという感じ．
無声を有声に誤る箇所が多くなる（非周期性指標でカバーするしかない）
また有声区間のF0の誤りのレンジこそ小さいので無視できるかも...?

もうちょい調べてみたら違うみたい
無声区間と評価された部分が大半占めてると死ぬっぽい(1600サンプルのみを使う場合)